Skip to content

scholarx

Universal Research Paper API + MCP Server + Agent for the agent-utilities ecosystem — one entry point for arXiv, PubMed Central, bioRxiv, medRxiv, PsyArXiv, OSF, and Semantic Scholar.

Official documentation

This site is the canonical reference for scholarx, maintained alongside every release.

PyPI MCP Server License GitHub

Overview

scholarx is a single, deduplicated interface across seven preprint and academic repositories. It fans a query out to every source in parallel, merges and deduplicates the results, downloads full PDFs, and ingests them into the epistemic-graph Knowledge Graph. The capability is exposed three ways:

  • ScholarXClient — a unified async Python client that fans out to all providers, applies per-source rate limiting, and returns merged, deduplicated SearchResult records.
  • Action-routed MCP tools — search, discovery, and storage tool modules, each togglable, that keep the tool surface compact for LLM contexts.
  • A Pydantic-AI graph agent (scholarx-agent) for autonomous, multi-step research workflows over the Agent Control Protocol and the Agent Web UI.

Deduplication is three-tier — DOI exact match, cross-identifier mapping, then fuzzy title and first-author matching — so the same paper appearing across multiple sources is collapsed into one record.

Explore the documentation

  • Installation — pip, source, extras, and the prebuilt Docker image.
  • Deployment — run the MCP server and the agent, Docker Compose, Caddy + Technitium.
  • Usage — the MCP tools, the ScholarXClient API, and the scholarx CLI.
  • Overview — ecosystem role, enterprise readiness, and architecture.
  • Concepts — the CONCEPT:SX-* registry.
  • Coverage Report — per-source coverage and verification.

Quick start

pip install "scholarx[mcp]"
scholarx-mcp                     # stdio MCP server (default transport)

Run a research scan from the CLI, or start the HTTP server:

scholarx scan --query "multi-agent orchestration" --output-dir ./papers
scholarx-mcp --transport streamable-http --host 0.0.0.0 --port 8004

See Installation and Deployment for the full matrix (PyPI extras, Docker image, all transports, the agent server, reverse proxy, DNS).